When free return shipping became the norm a decade ago, I suddenly found myself on a first-name basis with the folks that ran my local UPS store. Brands made it so easy for me to order a stack of new outfits in the mail, try them from the comfort of my bedroom, then put most of them back in the box and drop it off at UPS, FedEx, or the post office. At the time, e-commerce retailers were trying to lure consumers online by giving them an experience that was comparable—or better—than going to a store. Free returns, which made online shopping less risky for the customer, was a crucial part to that equation. But by 2018, 10% of all merchandise sold—or $369 billion worth of goods—was returned.
Free returns come at a huge expense to retailers. “There’s the logistical costs to think about, like the return shipping and warehousing,” says Manchit Madan, a data scientist at India-based e-commerce store Myntra Designs. “But then there’s the fact that when an item is in a customer’s home, they can’t sell it until the customer returns it. And then sometimes, when a product comes back, it is not in a re-sellable condition.”
That’s partly why many brands that first launched online—from Everlane to Warby Parker to Away—create brick-and-mortar stores. Allowing customers to see products in person and try them on makes it less likely that they will return them in the future. Now, Madan and his fellow researchers have developed a tool that shows how brands can cut down on the waste and cost of online returns. But consumers—particularly who love being able to order and return items freely—may not be happy with the potential implications.
Madan recently published a study that presents a way to predict the likelihood of customers returning a product—even before they have actually purchased it—with 83% accuracy. Along with his co-authors Sajan Kedia, who also works at Myntra, and Sumit Brar, who works at Google, he studied how consumers related to the more than half a million products that appear on Myntra’s website, along with the millions of orders and returns that roll in every week. They then built an algorithm that predicts the likelihood of an item being returned. They looked at everything from a customer’s shopping history to the specific items they put in their shopping cart.
They found some telltale signs that a customer will likely send a bunch of items back. For instance, returns often happen when customers fill their cart with many items. When a cart is filled with more than five items, that person has a 72% chance of returning an item, compared to 9% with one product. Older inventory—even at a discounted price—has double the return rate compared to newer items. And 4% of returns happen when there are similar products in a cart. “A person might have several different sportswear T-shirts, for instance, and they might be similar colors,” Madan says.
The larger narrative is that the data identifies different types of shoppers. Some people are in an exploratory phase, sampling a range of items and sizes before making a final selection. These customers treat online shopping much like a relaxed visit to a store, trying out different looks, and checking out different brands’ sizing. Other customers know exactly what they want and buy that item. These people might be reordering a product that worked for them in the past or replenishing their stock of items they have at home. “It’s not just about different kinds of shoppers,” says Madan. “It also has to do with the particular product they are shopping for. A customer might return a lot of products from one order, but not return anything from another.” (The National Retail Federation has tracked which categories of products have the most returns. After auto parts (22.58%), apparel has the next highest return rate (12.78%), followed by home goods (12.28%).)
For right now, this tool is currently only being used by Myntra. But the scientists have made their research available to other companies, which can adopt some of their ideas. The authors of this report suggest ways that brands can reduce the number of returns. Part of it has to do with intervening in situations where the customer seems confused or unsure about what they want. The researchers found that 53% of returns are due to size and fit issues, and if customers have multiple sizes in their cart, that’s a good signal that they don’t know what size they are with a particular brand. A retailer can prompt a customer to take a size survey or look at a size chart. That might also mean providing accurate sizing information in the first place. “With shirts, for instance, customers may not understand how the regular cut differs from the slim cut,” Madan says.
Then there are some customers who are just chronic returners. They like the freedom to order way too many items and decide what to keep later. Madan says there are ways to tame this behavior, too. For instance, the retailer can incentivize a customer not to return an item by offering a discount coupon at checkout, with the caveat that if they take it, they waive their right to a free return. Alternatively, and perhaps more controversially, a retailer could potentially vary the shipping costs in real time based on how likely the customer is to make a return. Brands don’t currently do either of these practices, but the authors present them as possibilities.
The latter solution, in particular, could generate backlash. Customers might feel unfairly targeted (or even punished) if a brand makes its shipping cost more expensive because they are likely to return items. And people might try to game the system to better understand how the algorithm tracks shipping costs. (Does buying five items at one go yield a higher shipping price than buying those five items separately?) Many customers have come to see free shipping as a right, rather than a privilege, and may feel resentful toward a brand that takes it away.
Madan says that variable shipping is just one tool at the retailer’s disposal, and there are many other things they can do to make returns less likely. In the end, repackaging and shipping back a product is inconvenient for customers, and reducing their need to make a return in the first place will improve their experience.